Research On EWMA Control Chart For Missing Data Process

With the increasing complexity of the production process,online algorithms in multivariate statistical process detection have been aroused more and more attention.However,conventional control methods on the assumption that the data is complete and obtained in equal intervals of time perform poorly when missing data exists.Considering using as much information as possible,we proposed an EWMA control chart for missing data process which is based on the weighting imputing approach that utilizes relationship between the complete ones and incomplete ones adaptively.In detail,it can monitor data under reasonable and comprehensive use of information,and accelerate the monitoring and alarm when data is out of control.The control chart proposed in this paper is fast and robust,the improved K-nearest neighbor interpolation value and the traditional EWMA interpolation value can be combined through our designed adaptive weighting function,and then the combined value will be used to impute the missing data in real-time,next we can monitor the complete data online.Among them,when there is a shift in a variable with missing value,a larger weight will be assigned to the EWMA imputation value adaptively,and vice versa.Besides,the optimal values of pre-specified parameters are suggested.Simulation results under Monte Carlo and an illustrative example of harmonic current monitoring of steel plate pressing machine demonstrate the robustness and sensitiveness of our proposed scheme.